From AI Adoption to AI ROI | Glean & Altimeter Capital | RAISE Summit 2026

Arvind Krishnan's Background and Core Philosophy

• Spent over a decade at Google, focusing on search, data separation, and the "needle in a haystack" information retrieval problem.
• Previous tenure at Rubrik emphasized enterprise data security, governance, and restricting agent access to authorized information.
• Translates these experiences to Glean by prioritizing high-quality information retrieval to ensure enterprise AI agents function accurately and securely.
• Views AI strategy as inextricably linked to data strategy, having established this foundation since Glean's founding in early 2019.

Market Position and Competitive Landscape

• Glean entered the market in early 2019, positioning the company as a pioneer before competitors like Microsoft Copilot, ChatGPT, or Anthropic addressed the specific enterprise search context gap.
• Competitors include Microsoft, OpenAI, and Anthropic, but Glean avoids re-innovating foundation models, instead partnering with leaders like Google, Anthropic, and OpenAI.
• Glean's competitive moat is the construction of the best enterprise "context graph" and retrieval technology, rather than building proprietary models.
• The company benefits from a "first-mover" advantage where market narratives often credit Glean as the pioneer even when Google launches similar products.

Adoption Trajectory and ROI Strategy

• Enterprise AI spending has shifted from initial adoption curiosity to cost containment, with some companies (e.g., Tesla) imposing strict token limits (e.g., $200/week) due to accelerating costs.
• Glean observed that personal productivity gains from AI are difficult to measure against business metrics, prompting a strategic pivot to departmental use cases.
• The company now focuses on high-value, measurable business processes such as customer service, RFP filling, and legal contract review.
• Example of ROI: A 40% increase in customer service resolution efficiency directly impacts operational budgets, allowing companies to fund further AI adoption without increasing headcount.
• Glean has evolved from a search tool ("Google for work") to a work assistant, and finally to a horizontal agent platform that provides context for other AI products.

Forward-Deployed Engineering (FDE) and Services Model

• Glean utilizes a limited FDE strategy to assist customers in building high-value agents and to learn patterns for product improvement.
• FDE revenue constitutes a very small portion of total revenue, distinguishing the approach from heavy professional services firms like Palantir.
• The long-term strategy involves partnering with consulting and services firms to scale implementation while Glean provides the underlying software platform.
• Krishnan emphasizes that successful AI delivery requires human intervention to drive business value, rejecting the idea of "magical" self-service adoption.

Model Provider Competition and the "Layers" Argument

• Glean maintains a neutral stance, integrating models from OpenAI, Anthropic, Google, NVIDIA, and open-source providers to prevent vendor lock-in.
• Krishnan predicts that within two years, the majority of enterprise inference workloads will shift to open-source models.
• The company positions itself in the layer above foundation models, specializing in integrating enterprise systems, data security, governance, and understanding organizational structure.
• Glean allocates more R&D resources to this enterprise context layer than any other company, viewing it as the critical differentiator for enterprise survival.
• The platform acts as a "climate controller" for CIOs, providing a stable environment regardless of which model providers ("seasons") dominate the market.

Technology Stance and Opinions

• AGI: Thumbs down; Krishnan expresses uncertainty regarding the definition and current relevance of Artificial General Intelligence.
• Forward-Deployed Engineering: Thumbs up; deemed essential for delivering tangible business value.
• Open Source AI: Big thumbs up; predicted to dominate enterprise inference in the near future.
• Model Context Protocol (MCP): Mixed review; initially praised for motivating SaaS companies to open APIs, but now criticized for creating over-reliance and inefficiency.
• Krishnan argues that relying on real-time MCP calls for every task increases AI costs and encourages "lazy" architecture, whereas pre-built context graphs are more efficient.
• Context Graphs: Identified as the key to driving high-accuracy results within enterprise environments.

Investment Outlook (Long/Short)

• Long: Bullish on all startups adding value to open-source language models, including those building inference stacks and surrounding infrastructure.
• Short: No specific companies; instead, Krishnan notes that the entire AI industry is currently under-supplied, delivering only ~5% of the feature set and value enterprises actually need.
• The only companies to avoid are those that doubt themselves or lack execution capability, given the massive gap between current technology supply and enterprise demand.